⚠ PRELIMINARY RESULTS — NOT FOR CITATION OR DISTRIBUTION ⚠
Analysis is ongoing and subject to revision. | Data: NIBRS 2021–2024 | N = 510,582 juvenile incidents

Overview

Column

Effect on Case Clearance Rates by State

State Profiles

Column

Select a State to Explore Its Results

⚠ Note on Nevada: NV’s ban took effect July 1, 2024 (~6 months of data). The lag plot shows an initial positive effect that reverses by month 4. NV estimates should be interpreted with caution.

State Law Details

Dynamic Effects

Column

All Crimes: Any Clearance Rate Over Time by State (pp)

How to read these plots: Each panel shows how the effect of the ban unfolded over time in a single state. The line tracks the estimated effect at each month after the law took effect. The shaded band is a 95% confidence interval — if it includes zero, the effect at that time point is not statistically distinguishable from no change.

Sex Crimes: Clearance Rate Over Time by State (pp)

Why sex crimes matter: Confessions play an outsized role in sex crime cases. If deception bans affect case outcomes, the impact may be most visible here. These panels show the same month-by-month analysis, focused specifically on sex crime clearances.

Exceptional Clearance

Column

Pooled Effect on Exceptional Clearance Subcategories

What this shows: Exceptional clearances are cases closed without an arrest — e.g., prosecution declined, victim refused to cooperate, or juvenile handled informally. Each dot shows the pooled average effect across all 7 states for that subcategory. Horizontal whiskers represent the pooled 95% confidence interval. Estimates whose CI does not cross zero are statistically significant.

Technical Details

Column

Randomization Inference

What this tests: If these bans had no real effect, how unusual is the observed result? I randomly re-assign treatment timing 500 times and re-estimate the pooled effect each time. The histogram shows this “null distribution.” If the actual estimate (vertical line) falls within the bulk of the distribution, the real effect cannot be distinguished from random noise.

Placebo Tests

What this tests: I check for “effects” at fake treatment dates (6, 9, and 12 months before the actual ban). If the research design is valid, these placebo effects should be zero — indicating no pre-existing trend that could be mistaken for a real effect. Points near zero (gray shaded zone) support the design.
Note: IL and OR lack estimates at 12 months before the ban due to insufficient pre-treatment data at that lead.

Bayesian Summary

Bayesian Posteriors

Agency-Level Effect Distributions: Arrest Clearance (pp)

What this shows: Each panel shows the distribution of agency-level treatment effects within a state. A distribution centered on zero means most agencies saw no change. Agency counts appear low because the analysis is restricted to agencies with enough juvenile incidents both before and after the ban to produce reliable estimates — most small or rural departments are excluded. States with shorter post-treatment windows (e.g., NV with ~6 months) have fewer qualifying agencies.